Audio Clock Locked to External Timecode?
 

When recording double-system sound, if the sample rates of the video camera and the audio recorder are different, when the files are combined and the heads lined up in post the audio will drift out of sync with respect to picture over the course of the shot. The documentation for the Tascam HD-P2 explicitly states that in addition to using its regular internal sample rate clock, it's able to slave the audio sample rate to the rate embedded in incoming timecode as well as to video blackburst, wordclock, and incoming digital audio. But does anyone know if other recorders -- I'm thinking here of the Sound Devices 702T/744T but also more genericaly -- can also be assumed to slave their audio sample clocks to the incoming sample rate when receiving external timecode? SD's manuals say they can slave their audio clock to wordclock or to the sample rate of incoming digital audio at the AES/SPDIF inputs but they don't mention slaving to the rate information embedded in external linear timecode. I know they'll slave their timecode clock to the external code but that's a whole different issue. Can one assume they also slave their audio sample clock to external code as well when they're in one of the external code modes? I've been unable to locate anything in their manuals or knowledge base that explicitly says yeah or nay. Or are the sample rate clocks in most DV cameras and quality recorders like the SD units sufficiently accurate that sync drift shouldn't be an issue? (And I know most prosumer cams don't have timecode output anyway but again, that's another issue.)

Hi Steve...

In my own experience, using a Sound Devices 702, syncing the head of the audio to the video via clapboard will maintain sync for roughly one hour. In about an hour, there is an error around 1-2 frames. I've dealt with this by doing a "stretch" of the audio without a pitch change...easy to do in Vegas. To the best of my knowledge, the Sound Devices "T" models will not easily allow a sync signal from a camera to force their internal timecode.

Yep - it was our earlier discussion about deriving wordclock from composite video that got me thinking about this. The way the Big Guys do it on location, especially for multicamera concert shoots, is to use something like Ambient Lock-It boxes. There's a Lock-It on each camera and one on the audio recorder. The cameras are genlocked to their Lock-Its while the audio recorder gets wordclock from its Lock-It. Ambient says sync drift will be less than one frame in 24 hours with this method when the Lock-Its are first tuned to the same master clock. But of course, as well as not outputting timecode not many prosumer grade cameras under the Canon XL-H1 accept genlock either, plus those Ambient boxes are not cheap! Man, Sony XDCAM is looking better and better <grin>!

Dear Steve,

I am not an expert in this area. However, I feel that word clock is not an issue for normal double system sound.

The Sound Devices 744T as well as the other "T" models use Ambient time code generation, which is very accurate, as well as tuneable for even more accuracy.

Case 1, If the timecode sync cable is left connected, both the camera and the recorder will be using the exact same timecode. One will be the master, the other will be the slave.

With the 744T either one can be the master. This is not true with the Tascam HD P1 recorder.

Case 2, If the timecode sync cable is only connected for syncing purposes, then they can drift, but for reasonable periods of time they will be right on.

Since it takes less than 10 seconds, and the 744T is so portable, it is very easy to resync at resonable intervals. Even if they are separated by a resonable distance, a BNC cable can usually be used for sync.

The 744T allows you to compare the internal timecode to the external timecode so that you can check how long the devices will remain in sync. Thus one can determine what a "Resonable Period" is for the equipment in use. In my opinion, it is more likely that the camera will drift than the 744T.

Now, the 744T also has word clock in and out. If you have two sources of sound, such as a 744T and another digital sound recorder, and you are combining the outputs of both in a live environment, then one would want to ensure that the exact start of each digital word (either 24 bits or 16 bits) are aligned. Connecting the "word out" signal of one device to the "word in" of another will accomplish this feat.

But, for normal double system sound, I assume that others do as I do. I record sound in the camera and sound in the 744T, so that it is easier to match them up in post and so I have a backup. I usually only use the 744T sound tracks for the sound in the final project. I may use the on-camera microphone for some ambient sound in the mix.

The important point is that I never play back sound from the camera and simultaneously play back sound from the 744T and mix or combine the signals together in any way in which I would need word clock output from the camera connected to word clock input of the 744T.

I assume that when the sound tracks are put into a non-linear editor (NLE), and the sound is played back, word clock is also not an issue since the NLE will have its own master clock which will be shared by all tracks and outputs.

It is up to the NLE editor to align the tracks, it is up to the NLE and the computer's sound card to ensure that the first bit of each word, for each track, start simultaneously.

I have an Ambient Slate, which is also a timecode generator and/or receiver. It has the same accuracy as the ambient timecode generator/receiver in the 744T. This means that I can use the Ambient Slate as the master, then temporarily sync to the 744T, then go around to all of the cameras and sync to them. This works very well. So after any breaks and or camera battery changes, I just resync all of the devices. Of course, depending on the devices, this may not be necessary.

I feel that syncing to blackburst would be helpful if the camera does not have timecode syncing capabilities. I assume that this requires a constant connection between the two. Please be advised that I have not studied blackburst in great detail.

I welcome any corrections to the above.

Dear Bill,

The 744T can easily be synced to an external timecode, either from a camera with timecode output such as a XL H1, or a good timecode slate, or to a master timecode generator.

There is a cable available from Sound Devices that plugs into the 744T using a 5-pin Lemo connection. The other end has two BNC connectors, one for timecode in, the other for timecode out. Then using the timecode memu, one can easily sync up the two devices with either one being the master.

Thanks for the reply, Dan. Good to see you again, by the way. Drop me an email and lets get caught up.

Timecode sync and sample clock sync are two different things. Jamming the SD recorder's timecode from an external source and/or jamming your smart slate from the recorder's timecode helps lining up - 'sync'ing' - sound and picture in post but it doesn't insure that they won't drift out of sync over a long shot. There's only one timecode reference actually recorded in a BWF file - the TC value of the first sample. Once sync is established, to stay in sync the recorder's sample rate must be such that when the camera records 2997 frames of video, the recorder should record exactly 4,800,000 samples of audio. If the audio clock is, for instance, somewhat faster, say 48.01kHz, that 2997 frames of video will have 4,801,000 samples of audio recorded at the same time. But in post, during the time 100 seconds of video is played, the NLE will only play 4,800,000 samples of audio instead of the 4,801,000 that were recorded. It is actually only playing the sound that was recorded in a little less than one second, the audio that was recorded in 4800000/4801000 or .99 seconds to be exact, and the two will eventually drift apart because the audio is playing slower than the video. And while the SD manuals clearly state they can slave their internal timecode generator to an external source such as the camera, what they don't say is whether doing so also slaves their sample rate clock at the same time. If they don't, and the audio sample clock is off, the timecode value coming from the camera as the first sample of audio is captured is 00:00:00:00 establishing sync but when the timecode value in both the camera and recorder gets to 00:00:01:00, the sound file will be writing the 48010th sample and not the 48000th sample as it should (using my example numbers). Of course my example is a very severe error to illustrate the point and I'm sure the SD sample clock is much more accurate than that.

By the way, wordclock sync insures that each each sample in the digital data comes from exactly the same point in time in both digital devices, allowing you to mix the two digital streams without burps, clicks, and pops. This is crucial when mixing music on a digital mixer, for example, but that's irrelevant for double system sound since you're not mixing a digital audio stream from the camera with more digital audio from the recorders A/D converters as you're recording anyway. (If you were using 2 744's to record a total of 8 tracks at once it would be another story.) BUT since it does sync the slave sample clocks to the master, a side benefit of wordclock is it will also prevent drift caused by slight differences in the two sample rates when the clocks are running independently.

Interesting to hear this info. Unfortunately, for me anyway, it's rather moot because my JVC HD110 has no word clock or Time Code I/O. Therefore, the only reference I could possibly use is the blackburst. That leaves me searching for a blackburst to worclock, portable converter.(BTW...cheap and battery powered ;o)

I've found one by Rosendahl that's about $700 US. Unfortunately it requires AC power it's really only suitable for a studio or similar indoor venue. They also make a unit that extracts TC from a LANC output but it too is AC powered only.

------------added
Wrote to Rosendahl this morning, telling them how badly the DV community is crying for a battery-powered product that combines the functions of their WIF and LIF products. We need a battery-powered device that is about the size of a Beachbox so it can fit on a camera's accessory plate, between a camera and tripod, or into a soundperson's kit, and will accept inputs from either LTC, LANC terminals with embedded LTC, blackburst, or composite video (with or without timecode) on the one hand and output stable SMPTE timecode and wordclock on the other. Curious if they'll respond and what they might say.

FYI all who are following this thread

Got a definitive answer from Sound Devices tech support this morning. In their 7xxT series recorders, the audio sample rate clock DOES NOT sync to an external timecode. It can lock to wordclock or the clock embedded in a digital audio stream presented to the S/PDIF input but otherwise it freewheels on its internal clock even though you're syncing the timecode to an external source.

Steve...

On your suggestion, I followed up on the Rosendahl website. Their BonsaiDrive is the perfect (well, almost perfect) solution for taking in video(black burst) and outputting wordclock for the SD 700 series audio recorders. It is exactly what I've been searching for, except for the price....$2000. The amazing thing about the Bonsaidrive is that it will RECORD 4:2:2 10-bit video on an 80 Gig hard drive via component inputs. That's pretty unusual. The caveat is that it only outputs(in addition to wordlock) component video, which means I'd need a BMD or Aja card to input the component signal into my computer.

Thank you for the follow through and update, Steve.